ClawHub
Back to skill

Security audit

AI KEY·标题与封面

Security checks for vulnerabilities and agentic risk

Overview

This title-writing skill also automatically changes shared writing rules and saves feedback without clear user control.

Review before granting write access. Require approval for shared-rule changes and make saved preferences optional and inspectable. Review the missing framework and companion documents separately; the advertised analysis script is not bundled. The empty static scan does not resolve the persistent-change concerns supported by the instructions.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

The skill mandates automatic updates to a shared writing framework, beyond its declared title-and-cover generation scope.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The skill explicitly directs the agent to automatically incorporate any user feedback into a shared writing framework, extending a title-generation task into persistent rule modification. This can promote task-specific or adversarial feedback into instructions affecting later tasks and other skills; personalization is a plausible benign purpose, but no approval or validation boundary is specified in the supplied file.

Content

No source excerpt is available for this finding.

The skill automatically persists user feedback without warning the user that title generation modifies local memory and rule files.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The skill requires automatic feedback persistence and separately instructs writing corrections into local memory, without requiring runtime disclosure or consent. A routine title request can therefore retain user-provided information and alter future behavior unexpectedly; local storage limits exposure compared with external transmission, but does not eliminate unwanted retention or persistent contamination.

Content

No source excerpt is available for this finding.